When Gemini first came out I thought, wow, it is Google so of course it will be good, and subscribed to AI Pro for a year.
The first two weeks or so were good. But as the days went by the quota policy kept changing at their whim, Gemini's nonsense kept growing, and after fixing and fixing my skills to make them usable, at some point it would spout nonsense again or make changes I never asked for; worn out by that, eight months passed with me fixing things by hand.
Now I have zero expectations even for 3.5 Pro, so after a persistent back-and-forth with Google I got a refund and moved back to Cursor after a long while.
I moved to Cursor Pro and have been using the default Cursor models for two days: Composer 2.5 standard and Grok 4.5, and these do better than the Claude 4.6 series that was included in GEMINI. The context is smaller than Gemini's, but I am not going to upload an entire project anyway, and even 200K or 250K produces results I do not need to touch. On the $20 plan I also tried sonnet5 a little, and it looks like it will last quite a while; after using it about 4 hours a day for two days I have used about 5% of the base model.
Gemini's quota usage has been erratic again lately, and I got caught by the weekly limit and had to wait 4 days, so I was quite worried since Cursor resets monthly, but at this level I am very satisfied.
Enough of you, Gemini.
